Claude Sonnet 5.5: nearly Opus 5.5 performance at half the price
September 28, 2026 · JICA LABS
On Monday, September 28, Anthropic launched Claude Sonnet 5.5, the second model in its 5.5 family, a week after Claude Opus 5.5. Sonnet is the company's everyday workhorse model, and this version comes very close to the flagship's performance without raising the price.
The key points
- Speed: it generates output more than 30% faster than Sonnet 5.
- Cost: API pricing is unchanged ($2 per million input tokens and $10 per million output tokens), but Anthropic says a complete task can cost up to 30% less, because the model uses fewer tokens and tool calls to finish it.
- Nearly Opus: on several tests it practically ties Opus 5.5, which costs twice as much per token.
- Safety: it is the first Sonnet to ship with cybersecurity safeguards like those of the more capable models. Higher-risk cybersecurity requests automatically fall back to Sonnet 5, and the app shows it.
The numbers
These are Anthropic's published results, compared with Sonnet 5 and Opus 5.5:
| Benchmark | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 (terminal coding) | 70.6% | 10.3% | 66.4% |
| OSWorld 2.1 (computer use) | 80.1% | 57.0% | 81.8% |
| CursorBench 4.0 (coding) | 55.5% | 34.1% | 57.8% |
| FrontierCode 1.1 (coding) | 46.2% | 42.4% | 54.4% |
| GDPval-AA v2.1 (knowledge work, score) | 1,844 | 1,449 | 1,846 |
The most striking jump is on Terminal-Bench: from 10.3% to 70.6%, even above Opus 5.5. Anthropic also says Sonnet 5.5 at low or medium effort beats Sonnet 5's best score at roughly a tenth of the cost per task. As always, these are tests chosen and run by the company itself.
What early customers report
According to cases published by VentureBeat, at Box it was more accurate and 2.4x faster while using 12% fewer tokens; at Zendesk tickets were processed 20% faster; and at Lovable it needed a third fewer tool calls.
Where it is available
In Anthropic's API as claude-sonnet-5-5, and on Amazon Web Services, Google Cloud and Microsoft Azure. In the Claude apps and Claude Code it defaults to medium effort. Anthropic confirmed that Haiku 5.5, built for high-volume, low-cost use, will arrive in the coming weeks.
Why it matters
Sonnet 5.5 now matches GPT-6 Sol on price ($2 and $10), OpenAI's mid-tier model. The race is no longer only about who has the most powerful model, but who delivers more for the same money. If you use these APIs, see our pricing comparison with a calculator, which now includes Sonnet 5.5.